Tuesday, April 15, 2008

Social Combat!

So, I've been dabbling in prototypes that test the idea of "social combat" - it's like normal combat, except it's not physical. The idea is to introduce the same level of complexity and depth to a social situation as you get in a combat situation in other games.

There have been a few attempts, but they're always pretty bad because they are either foolishly transparent or require some kind of wicked new AI.

Nothing makes this clearer than Oblivion, both with it's insipid social minigame and the fact that the haggling is simply based on your skill level.

It occurred to me:

Logical fallacies!

What if I made all the various techniques and stats related to various logical fallacies?

As I thought about it more, I became convinced it would work. Here's the overview:

The game is a trading game set in ancient somewhere. Egypt. Whatever. The idea is haggling. Various characters have various fallacies they are best at, and at higher skill levels (earned through play), they gain access to special, highly advanced comments of that type.

For example, if one of the characters is a street waif, she might have a specialty of "appeal to pity". She's an expert at making people feel bad for not giving her an excellent deal. As time goes on, she masters various specific techniques that allow her to choose between aggressive and passive approaches, between long-term benefit approaches or short-term approaches. For example, she learns the "tries so hard" technique, where you subtly get a low price, then pay a bit more because you "don't want them to feel pity for you". This technique wouldn't get you a very good deal, but sets up a long-term "hook" that can be used for even better deals at a later time from that same salesman. Depending on the situation, the salesman might decide, then and there, to refuse to take any money at all!

Like most fighting games, each kind of fallacy would have a different "elemental attunement". Appeal to pity, for example, is a simple "emotional" attunement, as is an appeal to ridicule or consequences.

Also, some techniques are better in certain situations. For example, ad hominem attacks aren't terribly useful straight-up. Yelling at the seller or buyer is useless. But you can use it to undermine their argument by insulting the person who gave them their information. It's a defensive technique, see?

Similarly, some techniques can reinforce your own techniques, working in concert. For example, if you're using an appeal to association: "most people like my apples, you'll like them, I'm sure!" You can add in an appeal to authority to give it more weight: "Even the queen likes my apples!" You can add in some ipsedixitism: "And apples are good for you!" plus an appeal to tradition: "Everyone knows that!"

Of course, then your buyer could retort with a bit of motive: "Of course you'd say that, you grow apples for a living!" Then attach some flattery: "It looks like you do a good job of it, too." Then a bit of special pleading: "How about you give me a sample? I've never had one before..."

Now, if the seller is immune to "emotion" damage, he negates both the motive comment and the flattery... but the special pleading is "almost logic", a different element. So a bit might get through...

What everyone wears is also important - specific clothes impart specific bonuses. Some give you some immunity to certain elements, others boost certain elements. Some offer protection against or strength in a certain fallacy type, others might actually only boost a specific technique within a fallacy...

What do you think?


Ryan said...

It basically sounds like the Pokemon battle system, with different names for attacks. (I'm talking about Red/Blue for the Gameboy. I hear it's different now, or something?)

"Starving Orphan Child! I choose you!"

I don't know... I think that in order to implement a fun "conversational battle system," you'd need to get away from rock-paper-scissors and hit points.

Craig Perko said...

Of course you do. I'm just trying to say it simply.

I've written a lot about it before. My take on the matter is that you can't have disposable encounters in a social situation. The idea is idiocy.

In this sort of situation, you'll be interacting with a lot of the same people over and over - your relationship to them is at least as valuable as how well you bargain today. After all, they'll order special for you if you're a good friend, and might refuse to sell to you at all if you're an enemy.

As I've said before, think of a person more as an equipment slot, rather than an encounter. You may put various things in the equipment slot over the course of the game, but you don't gain or lose equipment slots much. A person, like an equipment slot, is going to be with you for the whole game.

The problem is that, even implemented like that, it comes up shallow. The gameplay involved is too shallow.

By creating a paradigm in which complex gameplay can thrive - by porting over the basics of proven systems - you can fix that problem. And, even with disposable encounters, you'll end up with something playable.

Craig Perko said...

Here's a link:

Olick said...

This sounds like a really appealing idea. Wouldn't be easy to do with 100% realism (like in a video game), but it would be far more engaging and realistic than the options thus far.

I don't think that a numerical 'worn down' value would work for this, nor do I think he's suggesting this. In fact.. I'd say fuzzy logic would apply. Like a deal would have various states it could be in, which would represent whomever has the social (mental really) advantage here, and the change in the deal would only be noticeable in large increments. Like if in a situation, trying to appeal to his sense of nationalism might make him suspicious of you, lowering your status, however if first you've presented a 'good logical' argument, and already raised your status, he might not mind that you are ALSO nationalistic, because the small offense of nationalism is nulled by the strong appeal of logic.

I dunno, without getting into math its hard to describe, but it would be state-based, not point based. Maybe.

I guess long-term interactions would probably have to work in a different way as well.. I don't know how that would be modeled.

Christopher Weeks said...


I also like the notion that you might employ these tactics for not only long- vs. short-term gain, but also multi-dimensional purposes. Maybe by using my time during the encounter to plausibly flatter/befriend the person, you are not only getting a short-term worse result than is possible in exchange for future benefit, that benefit might not even be in the same domain as the current negotiation. (Since I'm not at all sure I stated that clearly, I mean e.g.: while haggling over apples with the orchard-keeper, and spending your "turns" making friends, the end result might not even be cheap apples -- it could be a hiding place when the invasion comes or support getting a seat on the Chamber of Commerce board, or whatever.

"Also, some techniques are better in certain situations. For example, ad hominem attacks aren't terribly useful straight-up. Yelling at the seller or buyer is useless."

If the game were multi-player, there's room for huge strategic manipulation/combinations. Good-cop, bad-cop is obvious. But maybe I use a ridiculously hard-line ad hominem during my encounter with someone in order to move them into a certain emotional state so that when you immediately follow up the encounter with one of your own, you have some kind of advantage -- that person is more susceptible to your friendly advances, etc.

Craig Perko said...

That's the sort of thing I was thinking of, yes. :)

Patrick said...

This reminds me of a (very old) post you wrote, I think it was from 2005, on transaction costs in social interactions. That's the fundamental thing, and the transactions don't nessecarily have to be merchant transactions. For instance, we barter with our friends and aquaintences all the time, for favor, for sex, for information, for bragging rights. If you got the dynamics right, you could do a lot.

The question of what mechanics would make that work smoothly is a harder one. Your idea of logical fallicies is appealing, it reminds me of the principle (rather than the explicit, technical function) of Godel's Theorem, where the appeal is an attempt to pose the undecidable proposition in the counterparty's logical frame, thereby creating a Black Swan event and tipping the results your way.

Of course, if you pose these ideas to a bunch of geeky males, they're going to clamor for social modeling based on transaction costs. Whatever prototype you come up with, be sure to playtest it with some women. And Hobos! Hobos are great playtesters. Women hobos, best of both worlds.

Tom Hudson said...

When you hit somebody with a sword, you see a gout of blood and hear an appropriate sound. When you hit them with an Appeal to Pity, what do you see/hear/read?

On a similar topic, most games give you some feedback for the current health of the monster you're fighting. How would you expose the current state of the social relationship you're modifying?

People have tried vaguely similar things in interactive fiction and not had much success; see Meta-Conversation Verbs at http://emshort.home.mindspring.com/NPC4.htm, or the end of the Ask/Tell section at http://www.tads.org/howto/convbkg.htm.

(Now, I like the idea, I'm even thinking about how to adapt it to a couple of prototypical situations, but I'm not sure how to overcome those problems myself.)

In Neverwinter Nights you could occasionally use persuasion and bluff skills, but those had to be manually written into the dialog trees - wouldn't seem to scale well up to large numbers of skills. In Oblivion, you had four different 'persuasion approaches' in a minigame, but there wasn't much content there.

Craig Perko said...

The problem isn't feedback, in my opinion. In most games - such as an RPG or an FPS - you get a response that is maddeningly vague when you attack. A number pops up over their head, maybe. They stagger a bit, maybe.

How effective you really were is usually not made clear. Through repeated battles, you eventually work out that a zombie has 30 HP and is weak to fire - or maybe you use a special object to tell you - and that allows you to reconstruct what is going on in your head even with the vague responses.

The way that people have tried it before - such as meta-conversation verbs - is fundamentally flawed.

I will explain why in a new post.